TikTok photo posts vs videos: what the engagement data says and why it conflicts

What do the two TikTok datasets report?

Both come from companies that sell social-media tools. Neither is peer reviewed. The table sets them side by side on the same lines.

The Fanpage Karma percentages above are the vendor's own wording. Buffer calls its report "not a full-platform view" and says images are "more viable than you might expect" on TikTok.

A third source adds noise rather than clarity. Metricool's 2024 TikTok benchmark press release, published on Social Media Today (opens in a new tab) on 18 November 2024, says video content receives more participation. A search summary of the same study says the opposite. We did not retrieve its sample or formula, so we do not use it.

Why can both datasets be honest and still disagree?

1. Different denominators. Buffer defines its engagement rate per reach in its Instagram figures, and we did not find the same statement for its TikTok figure. Fanpage Karma does not say what it divides by. A post with a small audience and loyal viewers scores high on a per-reach rate and low on totals.
2. Different averages. Buffer reports medians. Fanpage Karma does not state whether its figures are means or medians, and one breakout post can drag a mean far above a median.
3. Different samples. Buffer's posts were published through Buffer by accounts that posted at least 10 times in the past year across at least 4 weeks. Fanpage Karma does not describe its account mix.
4. Different periods. Five months in 2025 is not the same window as Buffer's data through December 2025.
5. Different labels. Buffer says "images" and Fanpage Karma says "carousels". Neither text says whether the two groups hold the same posts.

An example with fictional numbers shows how the first reason works. A photo post reaches 1,000 people and earns 40 engagements, which is 40 per 1,000 reached. A video reaches 10,000 people and earns 300 engagements, which is 30 per 1,000 reached. The photo post wins on rate and loses on total engagements. Both statements are true, and a dataset that reports only one of them can look like it contradicts a dataset that reports the other.

The averaging problem is just as easy to see. Take five videos with 300, 320, 340, 360 and 9,000 views. The mean is 2,064 views and the median is 340. Which one a study reports changes the story.

What does TikTok itself say about formats?

As of November 2026, TikTok's Help Center (opens in a new tab) says the For You feed draws on user interactions (likes, shares, comments, watching to the end or skipping, accounts followed), content information and user information. It adds that, for most users, interactions including time spent watching are generally weighted more than other factors. The page says nothing about weighting a photo post differently from a video. We read it in German, translated.

So there is no official ranking of formats to settle the vendor conflict. The honest summary is that the evidence does not support "photo posts beat videos" or the reverse.

How do you compare photo posts and videos on your own account?

Your account is the one dataset where the sample, the metric and the period are yours to control.

1. Write one question, for example: "On the same topic, do my photo posts earn more shares per 1,000 views than my videos?"
2. Pair the content. Make a photo post and a video on the same topic, posted close together at similar times of day.
3. Pick one metric and one denominator, and write both down. The numbers should come from the same place every time.
4. Read each post at the same age, such as 48 hours after posting.
5. Compare medians, not your best pair.
6. Decide in advance how many pairs you will collect, and stop there.

Here is a log of six pairs. The numbers are fictional and only show the layout.

The median is 22.5 shares per 1,000 views for photo posts and 17.5 for videos. The median is 1,050 views for photo posts and 2,450 for videos. The photo posts win in three pairs, the videos in two and one is a tie. On this fictional log the answer depends on the metric: photo posts do better per view, and videos collect more views. That is one way two honest datasets can disagree. Six pairs is too few to call it either way.

What can you not conclude from this data?

• That photo posts beat video on TikTok. The datasets conflict, and neither is peer reviewed.
• That video beats photo posts on TikTok. Buffer's result is a median engagement rate on Buffer-published posts only.
• A multiplier, such as the more than 81% higher engagement rate in Fanpage Karma's study (opens in a new tab), without the dataset name, the metric and the sample.
• That your account will behave like a vendor's median. Their denominators differ from yours.

For the wider evidence on formats, hooks and timing, see what makes slideshows perform, and for specific claims checked one by one, slideshow myths. The logging method turns the pair table above into a sheet you can keep. In Fyslide you can link a post and type in its numbers by hand (analytics guide); automatic analytics are Coming soon, included in your lifetime access as an update.